The Agent Skills Directory

[PROMPT_INJECTION]: The skill operates on user-provided code for refactoring, creating a surface for indirect prompt injection. 1. Ingestion points: The agent is directed to analyze code snippets or files provided by the user (as described in SKILL.md). 2. Boundary markers: The instructions do not define delimiters or "ignore embedded instructions" warnings for the processed code. 3. Capability inventory: The agent uses internal analysis and generation; no external tools are requested in the frontmatter. 4. Sanitization: There are no instructions for sanitizing or escaping the content being processed.
[COMMAND_EXECUTION]: The README.md file includes instructions for shell commands intended for skill development and validation. The documentation lists commands for node and pnpm (pnpm build, pnpm validate). One specific command targets a local script path: node ~/.claude/plugins/cache/dot-claude/dev-skill/*/scripts/validate-skill.js.
[EXTERNAL_DOWNLOADS]: The skill references several external documentation sources and libraries, all evaluated as safe under the trusted sources policy. References include reputable technical and educational domains such as Stanford University, Microsoft, Mozilla (MDN), Wikipedia, and official documentation for React and TypeScript. The rules also recommend using established open-source libraries like date-fns, lodash, zod, and p-limit for improved code quality.

same-results-less-code